Unreasonably sudden and sharp temperature increase

Dear LAMMPS users

simulation goal:
In LAMMPS, I’m studying the stability of a model under high temperatures (1000 K-2000 K) ; raising the temperature and observing the system reaction after relaxation. Once some certain number of bonds break, the assigned temperature is considered as the critical temperature of thermal stability.

ecnountered issue:
At some points, it was observed that the temperature increases severely to unbelievable numbers leading the system to blow off.

I supposed it was just some fluctations that needs to be controled, so I edited the fix npt by adding drag 0.5 as follows
fix 1 all npt temp ${stemp} ${stemp} 0.1 x 0 0 20 drag 0.5

However, that didn’t fix the issue.

I looked for similar cases in the Mailing List but they are different from this one.
Does anyone know why this could happen?

DP_A_DNT_20.data (140.7 KB)
DP_A_DNT_20.in (6.3 KB)
relaxed_DP_A_DNT_20.data (344.6 KB)
log.lammps (161.1 KB)

You use a time step of 1fs on a system containing hydrogen atoms. Didn’t I warn you about this before? This gives me a strong deja-vu feeling…

1 fs is far too large, especially at high temperature. Your time integration simply does not work reliably. Simple result of the GI-GO principle!

Using the “drag” keyword is another “band aid solution” that should be avoided at all cost. It is really only needed for incompressible bulk cystals. It will significantly alter you ensemble and thus diminish the value of your results.

Before running your heating experiment, you need to take your equilibrated system and run with fix nve instead of fix npt or fix nvt and monitor energy conservation at different time step sizes. The lighter the atoms, the noisier the potential (AIREBO is rather noisy) and the higher the temperature, the worse the energy conservation will get.
I would start with 0.25 femtoseconds, but you may need to go as low as 0.1fs.
This should be standard practice for any serious MD simulation and is discussed in any good text book on MD simulation (and is something that you should really care about).

1 Like

Never did. That made me suspect that I missed that from the topics and discussions with you and I reviewed all of them. You didn’t mention any of this

I can see that you give good support and valuable pieces of advice to everyone, however accompanied with sentences that make me and others feel offended, which is something I can also see happens to anyone else here you reply to. Thank you for your kind help and for reminding me about what I should care about.